Presented 3 pairs of letters to 26 college-age ss in a 2-choice verbal discrimination paradigm. Each pair had an independent probability of being correct (pi). Within a pair these values were identical, while between pairs the reward probabilities differed. During training, ss were given feedback only on the cue chosen. The training trials were followed by nonrewarded (no information) transfer tests consisting of all combinations of the letters. No preference was observed during training, but the transfer tests revealed a preference for the cue with the higher pi value in training. An information model (in which it is assumed that ss make inferences about outcomes associated with the unchosen cue of a pair but that these outcomes are less effective than directly received outcomes), and a variant of a "scanning model" were supported. (PsycINFO Database Record (c) 2006 APA, all rights reserved).
- multiple-cue probability training & subsequent selection of cues associated with distinct reward probabilities